Learning to play in a stylized (chinos) game: some preliminary results

摘要

In a previous paper, we studied the rich strategic and information-revelation considerations arising in simple guessing game which is very popular in some European countries: the so-called Chinos game. Here, we complement that analysis with a preliminary study of the learning dynamics induced when players no longer can assume that others play according to fixed (possibly equilibrium) rules but, instead, have to use past evidence to understand or interpret their opponent's behavior. We show that if all players are flexible learners, they will eventually succeed in playing according to the unique Nash equilibrium of the game. This outcome, however, is also found to be rather fragile: if only one player (say, a "noise agent") is unable or unwilling to learn, this wrecks havoc in the overall population dynamics. In that case, this dynamics never settles and a complex cyclical pattern typically ensues. The implications of this conclusion for the study of information-revelation mechanisms in the real world (e.g. asset markets) is quite apparent but yet to be tackled in full detail.